Tensorflow based Deep Q Network
High Correlation Data Removing Method for Deep Reinforcement Learning in Obstacle Avoidance and Path Planning
In this paper, we propose a High Correlation Data Removing Method for Deep Reinforcement Learning, especially for Deep Q Network, which combines Neural Network with Q-learning. The proposed method can reduce the training time in following ways. Firstly, remove the unnecessary repetition and give priority to some specific actions. Secondly, store little high correlated transitions. At last, learn after a specific number of steps of storage.